1 / 26

Converter Design

Converter Design. Design Review, October 31, 2007 Josh Myers, Richard McCarthy, Paul Schwinberg, Daniel Sigg,. Initial LIGO Limitations. Collocation of analog and digital Signals converge at converter nodes Limited timing support

dinesh
Download Presentation

Converter Design

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Converter Design Design Review, October 31, 2007 Josh Myers, Richard McCarthy, Paul Schwinberg, Daniel Sigg,

  2. Initial LIGO Limitations • Collocation of analog and digital • Signals converge at converter nodes • Limited timing support • Clock input only, no synchronization input, Rely on cycle count • Poor noise performance • Sophisticated whitening & dewhitening filters, switchable • Inadequate throughput • VME bus latency limitations, 4 boards max. • Convoluted data paths • Network topology grew over time, Reflective memory limitations • High costs • Dependent on a single manufacturer Advanced LIGO

  3. Requirements • Timing • Absolute timing precision relative to UTC: 1μs • Guaranteed (no counting of cycles forever) • Latency between converter and processor: < 5μs • Noise • Converter range: ±10V SE or ±20V DE • Converter noise: 100–300nV/√Hz (SE, best effort) • No electrical connection between converters and processors (fiber) • Others • No restriction on converter location • Detection of transmission errors • Support for diagnostics Advanced LIGO

  4. State of Technology • SD-Modulators • Typically paired with FIR filters, long delays • SAR (Successive Approximation Register) • Analog Devices AD7634: 18 bit, 570 kHz, SNR 101 dB • Advanced Segment • TI PCM1794A, 24 bit, 200 kHz, SNR 132 dB (A-weighted / fNyquist~12kHz) • Sweet spot around ½ MHz sampling rate Advanced LIGO

  5. PCM1794 16 bit AD7634 PCI66-16AI64 Pentek ADC Advanced LIGO

  6. Comparison Advanced LIGO

  7. ADC Performance Advanced LIGO

  8. DAC Performance Advanced LIGO

  9. In-House System CPU Sensor Whitening ADC FPGA No Anti-Aliasing Boards No Anti-Image Boards Strongly Reduced Whitening Reduced Dewhitening analog analog digital fiber Actuator Dewhite DAC FPGA AA AI Advanced LIGO

  10. Whitening/Dewhitening • Compress the analog signal to fit the digital SNR AS port whitening Coil driver dewhitening Pentek ADC → D060535: Win 45 dB! FDI DAC → D060293: Win 10-20 dB Advanced LIGO

  11. Integrated Timing • Converter Clock • VCXO locked to timing receiver • FPGA clock & converter clock are the same • Synchronization • 1 PPS signal from timing receiver • FPGA counters are synchronized by 1 PPS • Data Transfer • Full time stamp on both send and receive • Sender deterministic • Receiver time stamp checked against allowed time interval Timing is guaranteed in hardware! Advanced LIGO

  12. ADC Board (D060535) • 16 channels input • Analog front-end • Fully differential, 40 Vpp • 5th order Cheby, 200kHz, 0.5dB ripple • 524288 Hz sampling rate • Analog Devices AD7634 • Onboard voltage regulators & reference • LVDS interface to FPGA Advanced LIGO

  13. Regulators Channel 1 Reference LVDS Input EEPROM LVDSTranslators Channel 16 Advanced LIGO

  14. DAC Board (D060293) • 16 channels output • Analog front-end • Fully differential, 40 Vpp • 5th order Cheby, 53kHz, 0.1dB ripple • 524288 Hz sampling rate (64 kHz bandwidth) • Texas Instruments PCM1794A • Onboard voltage regulators & reference • LVDS interface from FPGA Advanced LIGO

  15. Channel 16 Reference EEPROM LVDSTranslators LVDS Output Regulators Channel 1 Advanced LIGO

  16. Low-Drop Low-Noise Voltage Regulators Advanced LIGO

  17. Crate & Backplane • Eurocrate IEEE 1101.1/1101.10 • 6U x 280mm x 6HP • 2 converters per controller • Connectors • VME type for converters • VME64X type for controllers • All connections are point-to-point (no bus) • 1 unit = 2 controllers + 4 converters • ±16.5V, ±6.5V & ±24V analog supplies • 12V digital supply → 5V & 3.3V digital 1 unit Advanced LIGO

  18. Controller Board • Xilinx Spartan 3A DSP: XC3SD1800A (2x) • 37000 logic cells, 84 DSP slices, 1.5 MBit RAM • Converter control: clocks and serial interfaces (LVDS) • Filter Engines • Time-multiplexed serial links to uplink FPGA • Xilinx Virtex 4FX: XC4VFX20 • 2 x gigabit ethernet SFP transceivers & EMACs • Timing transceiver & logic • AES/EBU interfaces for mixed analog-digital testing • Digital IO lines • Digital power supplies • 12V input to 5V, 3.3V, 2.5V, 1.8V, 1.5V, 1.2V & whatever • Multi-phase synchronous Advanced LIGO

  19. Backplane Links Advanced LIGO

  20. Filter Engine (1) • Multiple second-order sections: • Formula for a single SOS: • 4 multiplications with coefficients, old input and old output values • 4 accumulations • c0 is a shift operation Advanced LIGO

  21. Filter Engine (2) Advanced LIGO

  22. Filter Engine (3) 6th order elliptic low pass900 Hz cut-off 35 bit resolution / 18 bit input 52 bit resolution / 18 bit input 52 bit resolution / 27 bit input Advanced LIGO

  23. ADC & DAC Performance • Josh Myers G070604-B Advanced LIGO

  24. Development Status • ADC board • Prototype and testing done • Ready for production with small revisions • DAC board • Prototype in hand • Testing in progress • Controller board • Schematics done • Filter engine has been simulated and tested • Simulations for uplink code are underway • Crate • Backplane defined • Thermal loading under investigation • Power supplies available as prototype Advanced LIGO

  25. Development Plans • Finished testing of DAC board • End of 2007, $7000 if another revision is required • Build and test controller board • Manufacturing by Dec 2007, testing by March 2008 • Simulation models by Jan 2008 • $14000 (two revisions) • Crate • Build backplane by Jan 2008, $4000 • Assemble prototype crate by Feb 2008, $5000 • Integration • Testing with computer back-end by mid 2008, $2000 • Small production run for testing at LASTI, 40m, etc., $56000 • Test stand for verification and automatic testing • Requires outside help, $13000 Advanced LIGO

  26. Conclusions • Highest performance ADC and DAC boards • Fully integrated timing and synchronization • Computer doesn’t see oversampling • Works at 16384Hz and 2048Hz • Compute load at these frequencies shouldn’t be a problem • Design owned by us • No dependency on single board manufacturer • Person power for in-house development & testing required • Clear path for future extensions and upgrades • Good reasons for adopting this system as the new baseline Advanced LIGO

More Related